From 97cd2c41ade8a9c9d85e40299ef4e3bbdddc7903 Mon Sep 17 00:00:00 2001 From: Yongye Zhu Date: Wed, 6 May 2026 03:14:49 +0000 Subject: [PATCH] fix precommit Signed-off-by: Yongye Zhu --- vllm/v1/attention/backends/mla/prefill/tokenspeed_mla.py | 4 ---- 1 file changed, 4 deletions(-) diff --git a/vllm/v1/attention/backends/mla/prefill/tokenspeed_mla.py b/vllm/v1/attention/backends/mla/prefill/tokenspeed_mla.py index 809d0bbddfd..135e76795b3 100644 --- a/vllm/v1/attention/backends/mla/prefill/tokenspeed_mla.py +++ b/vllm/v1/attention/backends/mla/prefill/tokenspeed_mla.py @@ -70,8 +70,6 @@ class TokenspeedMLAPrefillBackend(MLAPrefillBackend): qk_rope_head_dim: int, v_head_dim: int, vllm_config: "VllmConfig", - device: torch.device, - layer_names: list[str] | None = None, ) -> None: super().__init__( num_heads=num_heads, @@ -81,8 +79,6 @@ class TokenspeedMLAPrefillBackend(MLAPrefillBackend): qk_rope_head_dim=qk_rope_head_dim, v_head_dim=v_head_dim, vllm_config=vllm_config, - device=device, - layer_names=layer_names, ) # Pre-JIT the kernel for both BF16 and FP8 prefill shapes so the first